Reply to thread

Message

<blockquote data-quote="stinger78" data-source="post: 1082661" data-attributes="member: 6771">I think the risk is the early bump (not entirely undeserved, but certainly not understood) from their early OOC wins - particularly against a subpar ACC with many rebuilding teams. You can’t know all that up front, thus the bump. That’s OK but it must be addressed and corrected.Look at the RPI from last season at 2-weeks periods. You’ll see it. There were 4/5 SEC teams in the SOS top 25 initially. That’s about right. It stayed 5/6 past NYD. Again, based on OOC results, not bad. But then by the end of Feb there were 9 SEC teams in top 25 SOS. How did that happen? They played only themselves and the feedback loop took over, IMO.Anytime your output from one iteration of an algorithm is then fed back into it as input for the next iteration you risk this. A moderating feature should be added that senses and responds to dampen the feedback. It’s basic complex system dynamics: probe, sense, respond. You never know when a feedback loop might arise, so you have to probe for them and respond accordingly.As it were, the NCAAT provided that moderating input. However, the bloated SOS numbers had already justified 14 of 16 SEC teams getting slots, several with horrible conf records. The only way to justify that, again IMO, was the hyped SOS ratings that weren’t corrected.Not saying it was surely human bias (though there likely was), but it was the data error that was (apparently) not corrected prior to its use by the committee.</blockquote>

[QUOTE="stinger78, post: 1082661, member: 6771"] I think the risk is the early bump (not entirely undeserved, but certainly not understood) from their early OOC wins - particularly against a subpar ACC with many rebuilding teams. You can’t know all that up front, thus the bump. That’s OK but it must be addressed and corrected. Look at the RPI from last season at 2-weeks periods. You’ll see it. There were 4/5 SEC teams in the SOS top 25 initially. That’s about right. It stayed 5/6 past NYD. Again, based on OOC results, not bad. But then by the end of Feb there were 9 SEC teams in top 25 SOS. How did that happen? They played only themselves and the feedback loop took over, IMO. Anytime your output from one iteration of an algorithm is then fed back into it as input for the next iteration you risk this. A moderating feature should be added that senses and responds to dampen the feedback. It’s basic complex system dynamics: probe, sense, respond. You never know when a feedback loop might arise, so you have to probe for them and respond accordingly. As it were, the NCAAT provided that moderating input. However, the bloated SOS numbers had already justified 14 of 16 SEC teams getting slots, several with horrible conf records. The only way to justify that, again IMO, was the hyped SOS ratings that weren’t corrected. Not saying it was surely human bias (though there likely was), but it was the data error that was (apparently) not corrected prior to its use by the committee. [/QUOTE]

Verification

Who was the starting QB of the 2024 season?